jofi˙1597
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
JOFI
jofi˙1597
June 25, 2010
Dispatch: June 25, 2010 CE: AFL
Journal MSP No. No. of pages: 35
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
19:56
PE: Beetna
THE JOURNAL OF FINANCE • VOL. LXV, NO. 5 • OCTOBER 2010
Short Sellers and Financial Misconduct
JONATHAN M. KARPOFF and XIAOXIA LOU∗
ABSTRACT
We examine whether short sellers detect firms that misrepresent their financial statements, and whether their trading conveys external costs or benefits to other investors.
Abnormal short interest increases steadily in the 19 months before the misrepresentation is publicly revealed, particularly when the misconduct is severe. Short selling
is associated with a faster time-to-discovery, and it dampens the share price inflation
that occurs when firms misstate their earnings. These results indicate that short sellers anticipate the eventual discovery and severity of financial misconduct. They also
convey external benefits, helping to uncover misconduct and keeping prices closer to
fundamental values.
SHORT SELLING IS A CONTROVERSIAL ACTIVITY. Detractors claim that short sellers undermine investors’ confidence in financial markets and decrease market
liquidity. For example, a short seller can spread false rumors about a firm
in which he has a short position and profit from the resulting decline in the
stock price.1 Advocates, in contrast, argue that short selling facilitates market
efficiency and the price discovery process. Investors who identify overpriced
firms can sell short, thereby incorporating their unfavorable information into
market prices. In his account of short selling in Allied Capital, Inc., hedge fund
∗ Foster School of Business, University of Washington and Lerner School of Business and Economics, University of Delaware. We thank especially Jerry Martin, who maintains the KarpoffLee-Martin database used in this study, and also Anup Agrawal, Uptal Bhattacharya, Hemang
Desai, Karl Diether, Avi Kamara, Adam Kolansinski, Jennifer Koski, Srinivasan Krishnamurthy,
Paul Laux, Paul Malatesta, Charu Raheja, Ed Rice, Ronnie Sadka, Mark Soliman, Ingrid Werner,
two Journal of Finance referees, the Associate Editor, Campbell Harvey, and seminar participants
at the 2008 CRSP Forum, Concordia University, Yale Law School, Binghamton University–SUNY,
Rutgers University, Syracuse University, Temple University, University of Indiana, University of
Washington, Vanderbilt Law School, and the California Corporate Finance Conference for helpful
comments. We also thank the Q Group, The CFO Forum, and the Foster School of Business for
financial support.
1 There are many anecdotes about such strategies, which former SEC Chairman Christopher Cox
called “distort and short” (see “What the SEC really did on short selling,” The Wall Street Journal, July 24, 2008, A15). In 2000, for example, investor Mark Jakob turned a $241,000 profit
by shorting Emulex stock and spreading an internet rumor that Emulex’ CEO was stepping
down amid an SEC investigation (see http://www.sec.gov/litigation/litreleases/lr16747.htm and
http://www.sec.gov/litigation/litreleases/lr16857.htm). Leinweber and Madhavan (2001) report a
case in which investors shorted Sea World stock and spread false rumors that Shamu, Sea World’s
main attraction, was ill. For other examples, see Alistair Barr, “Short sellers: The good, the bad
and the ugly,” MarketWatch, June 13, 2006.
1879
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1880
June 25, 2010
19:56
The Journal of FinanceR
manager David Einhorn argues that short sellers even help uncover financial
reporting violations (Einhorn (2008)).2
In this paper, we investigate whether short sellers do in fact identify overpriced firms, and whether in the process they convey external benefits or harm
upon other investors. To do so, we measure short selling in a set of firms that,
ex post, clearly were overpriced: those that are disciplined by the SEC for financial misrepresentation. In our sample of 454 firms from 1988 through 2005,
96% have negative abnormal returns on the days their misconduct was publicly revealed, with an average 1-day stock price decline of 18.2%. These firms
therefore provide a natural test of the view that short sellers can anticipate
bad news.
The results of three tests indicate that short sellers are proficient at identifying financial misrepresentation before it becomes public. First, abnormal
short interest rises significantly in the 19-month period before the misrepresentation is publicly revealed. Second, the amount of short selling is positively
related to measures of misconduct severity, indicating that short sellers take
larger positions when the misrepresentation is particularly egregious. And
third, short interest-based indicators of financial misrepresentation in any
given firm-month are significantly related to the actual presence of misrepresentation, as revealed in subsequent SEC documents.
We also investigate whether short selling has external effects on other investors. We do not find evidence that short selling imposes external harm by
triggering a cascade of selling when the misconduct is publicly revealed. To
the contrary, short selling conveys positive externalities to other investors, in
two ways. First, the amount of prior short selling is positively related to how
quickly the misconduct is publicly revealed. Our point estimates indicate that,
among firms that are 12 months into their misrepresentation, those with abnormal short interest at the 75th percentile will be publicly revealed 8 months
before firms at the 25th percentile.
Second, short selling dampens the amount by which prices are inflated while
firms report incorrect financial statements. This improves price efficiency and
decreases the transfer from uninformed investors who buy shares from insiders
or the firm before the misconduct is revealed to the public. We estimate that
this price impact translates into savings for uninformed investors of around
1.67% of the firm’s market capitalization on average. Some of these savings
are captured by short sellers, who earn profits that average 0.58% of equity
value. Even net of such profits, the average net external benefit to uninformed
investors equals 1.09% of the firm’s equity value.3
These findings do not address whether short selling in general is informed
and beneficial for other investors. For example, we cannot rule out the possibility that some short sellers are noise traders, or that some seek to manipulate
prices through false rumors. But in our events—in which company managers
2 Lamont (2004) and Jones and Lamont (2002) summarize the debate over whether short selling
fosters market efficiency or facilitates harmful manipulation. See also Wilchins (2008).
3 These point estimates correspond to our first measure of abnormal short interest, ABSI(1).
Depending on the specific measure, our point estimates of the net external benefit range from
0.19% to 1.53% of equity value. See Section V.C and Table IX below.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1881
produce falsified financial statements—short sellers play a significant role in
identifying, uncovering, and mitigating the effects of financial misconduct.
This paper is organized as follows. In Section I, we review related research
and argue that our sample and test design are uniquely well suited to examine whether short sellers anticipate and help uncover financial misconduct.
Section II describes our data and measures of abnormal short interest.
Section III reports on tests of short sellers’ ability to anticipate financial misconduct, and Section IV examines short sellers’ external effects on other investors.
Section V concludes.
I. Related Research
Our investigation is related to a large body of research that examines whether
short sellers target overvalued stocks.4 The results are somewhat mixed.
Asquith and Meulbroek (1996) and Desai et al. (2002) find that stocks that
are highly shorted in 1 month tend to underperform in the next month, and
Diether, Lee, and Werner (2009) find that short sellers appear to take advantage of short-term overreaction in stock prices. Christophe, Ferri, and Angel
(2004), Christophe, Ferri, and Hsieh (2009), and Liu, Ma, and Zhang (2008)
find that short selling increases before negative earnings announcements, analyst downgrades, and mortgage loss-related write-downs. In contrast, Daske,
Richardson, and Tuna (2005) do not find any predictive ability of short selling,
and Henry and Koski (2010) find no evidence of informed short selling around
SEO announcements.
Our empirical tests employ measures of abnormal short interest that condition on firm characteristics, and thus are related to inquiries into whether short
sellers use information about firm fundamentals. Dechow et al. (2001), Asquith
et al. (2005), and Duarte, Lou, and Sadka (2006) find that short interest is related to market capitalization, book-to-market, and momentum. Richardson
(2003) fails to find evidence that short sellers target firms with high accruals.
But Cao, Dhaliwal, and Kolasinski (2006) find that short sellers do target firms
with high accruals after controlling for surprises in earnings announcements.
We find that short interest is related to accruals, as well as market capitalization, book-to-market, momentum, insider selling, institutional ownership, and
share turnover.
Three prior studies are most closely related to ours. Dechow, Sloan, and
Sweeney (1996) report an increase in short interest in the 2 months before
an SEC release in a sample of 27 Accounting and Auditing Enforcement Releases. Desai, Krishnamurthy, and Venkataraman (2006) and Efendi, Kinney,
and Swanson (2006) examine short selling before the accounting restatements
in a database compiled by the Government Accountability Office (GAO).5 Our
investigation differs from these papers in several ways. First, we introduce
4 See Figlewski (1981), Asquith and Meulbroek (1996), Desai et al. (2002), and Asquith, Pathak,
and Ritter (2005).
5 For a description of the GAO restatement data, see http://www.gao.gov/special.pubs/gao-061079sp/.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1882
June 25, 2010
19:56
The Journal of FinanceR
several controls for the severity of the misconduct, allowing us to infer whether
short selling affects stock prices directly, or whether it merely serves as a proxy
for misconduct severity. Second, we examine whether short selling tends to concentrate in the misconduct firms. And third, we estimate the external effects on
uninformed investors—including whether short selling helps expose financial
misconduct and whether it dampens price inflation during the violation period.
The Internet Appendix contains a tabular summary of the results that are new
to this paper.6
The data we use also provide for more powerful tests than the GAO restatement data. Hennes, Leone, and Miller (2008) report that 76% of the restatements in the GAO database are simple errors rather than misrepresentation
or fraud, a concern also expressed by Files, Swanson, and Tse (2009). This suggests that the GAO database contains a large number of misclassified events.
Even when restatements do reflect financial misconduct, they can occur many
months after the misconduct is public knowledge. In our sample, SEC inquiries
into financial misconduct are resolved 41 months after the initial public revelation, on average. Using a restatement that is made during or after that
41-month period would misclassify when the misrepresentation was or was
not public knowledge.
II. Data and Short Interest Measures
A. Financial Misrepresentation Data
To avoid the data problems discussed above, we use the Karpoff, Lee, and
Martin (2008a, 2008b) (hereafter KLM) database to identify all 632 SEC enforcement actions for financial misrepresentation initiated from 1988 through
2005.7 These data identify the period during which the misrepresentation occurred and also the trigger event, which is the initial public revelation of the
misconduct. This allows us to focus on short selling around the initial public
revelation. Short interest data are available for 474 of the 632 firms, and 454
firms have sufficient data on CRSP to calculate returns on their revelation
dates.
To illustrate the nature of our data and tests, it is useful to review the sequence of events that constitute an SEC enforcement action.8 These events are
summarized in Figure 1. Most enforcement actions follow a conspicuous trigger
6 The Internet Appendix is available on The Journal of Finance website at http://www.afajof.
org/supplements.asp.
7 Karpoff et al. (2008a, p. 10) report that the database is collected from “ . . . Lexis-Nexis’ FEDSEC:SECREL library for information on SEC securities enforcement actions, the FEDSEC:CASES
library for information on litigated enforcement actions, and the Academic Business News, General
News, and Legal Cases libraries for news releases (frequently issued by defendant firms) about
each enforcement action . . . the SEC’s website at http://www.sec.gov, which contains all SEC public
releases relating to enforcement actions since September 19, 1995. . . the Department of Justice
itself, which provided . . . further data on enforcement outcomes [, and] the Department of Justice’s
Corporate Fraud Task Force website at http://www.usdoj.gov.”
8 The following two paragraphs follow Section III in Karpoff et al. (2008b).
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1883
Enforcement Period
Regulatory Period
Violation Period
Violation
Begins
Violation
Ends
*
Trigger
Event
Violation to Revelation
Inquiry Investigation
Event
Event
Wells
Notice
Proceedings Events
Initial
Concluding
Regulatory
Regulatory
Proceeding
Proceeding
Enforcement Events
* The initial filing of a private lawsuit usually occurs soon after the trigger event.
Figure 1. Timeline of a typical enforcement action.
event that publicizes the potential misconduct and attracts the SEC’s scrutiny.
Common trigger events include self-disclosures of malfeasance, restatements,
auditor departures, and unusual trading. Here are two examples of trigger
events from our sample:
1. On November 21, 2000, Lucent Technologies Inc. announced that it had
identified a revenue recognition issue in its already-reported fourth quarter report as the company was completing its financial statements for fiscal
year 2000. The company also told investors not to rely on its first-quarter
forecast of 2001. Share prices fell 16% on the announcement day.
2. On November 13, 2003, Virbac Corporation announced that it was delaying the filing of its third-quarter 2003 Form 10-Q. Share prices fell 22%
on the announcement day.
Following a trigger event, the SEC gathers information through an informal
inquiry that may develop into a formal investigation of financial misconduct.
At this point the SEC may drop the case, in which case it does not appear in our
sample. If the SEC proceeds, it typically sends a Wells Notice to prospective
defendants, notifying them that it intends to begin enforcement proceedings.
It then imposes administrative sanctions and/or seeks redress through civil
actions. Some cases are referred to the Department of Justice and lead to
criminal charges as well. The SEC releases its findings and penalties in its
Administrative Releases and Litigation Releases, and every enforcement action
in our sample has at least one such release. These releases provide detailed
information on the period over which the misrepresentation occurred—which
we label the violation period—as well as other information that we use in our
empirical tests.
As reported in Table I, the events illustrated in Figure 1 typically take several
years to play out. In our sample of enforcement events, the median length of
the violation period is 24 months, and the median length from the beginning of
the violation until its initial public revelation is 26 months. The period from the
initial public revelation until the end of the enforcement action is an additional
41 months. Table I shows that the number of enforcement actions, the median
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
The Journal of FinanceR
1884
Table I
Description of the Financial Misrepresentation Sample
This table describes the yearly distribution of the 632 SEC enforcement actions for financial
misrepresentation from 1988 through 2005. The violation period is the date the financial misrepresentation began until it ended, as identified in SEC litigation or administrative releases. The
revelation date is the earliest date that information about the misrepresentation was made public. Most revelation dates are identified in SEC releases, and the remaining are identified in the
Karpoff et al. (2008a, 2008b) database. Revelation events include firm disclosures, restatements,
auditor changes, SEC filing delays, whistle-blower charges, class action lawsuit filings, bankruptcy
filings, and SEC actions (informal inquiries, formal investigations, Wells Notices, or first regulatory
proceeding).
Violation Period
(Months)
Violation Beginning
to Public Revelation
(Months)
Year
Number of Cases
Mean
Median
Mean
Median
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
25
13
28
34
35
32
46
29
37
34
36
36
67
49
70
31
17
13
20
25
20
34
26
24
22
26
29
26
30
33
27
26
32
32
32
42
24
23
15
24
24
20
17
24
27
24
21
30
24
21
24
33
32
36
31
30
29
35
32
31
29
28
33
32
33
33
28
25
33
35
41
46
28
23
22
30
28
24
22
25
27
24
25
24
23
20
27
29
36
36
Total
632
28
24
31
26
violation period, and median period from the beginning of the violation to its
public revelation generally increased from 1988 to 2005.
Panel A of Table II reports that news about financial misrepresentation is
associated with large declines in stock price. Return data are available for 454
of our sample firms. For 359 of these firms, the trigger event reported in the
KLM database is identified in the SEC’s administrative and litigation releases.
The mean 1-day market-adjusted return on the SEC-identified trigger date is
−20.7%, and the median is −15.0%.
For 95 of our events the SEC identified no trigger date, or the KLM database
indicates that there was an earlier public revelation of the misconduct. In 37
of our events, for example, the start of a class action lawsuit is the earliest
public revelation of the misconduct. The mean 1-day market-adjusted return
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1885
Table II
Share Price Reactions to Announcements of Financial
Misrepresentation
This table presents summary statistics on the 1-day market-adjusted returns for key dates in the
sample of 454 SEC enforcement actions for financial misrepresentation from 1988 to 2005, for which
sufficient returns data are available on CRSP. The market-adjusted return is the firm’s return
minus the CRSP value-weighted return on the same day. Panel A reports market-adjusted returns
for the revelation date, which is the earliest date that information about the misrepresentation
was revealed to the public. Most (359) revelation dates are identified by the SEC, and include firm
disclosures, restatements, auditor changes, SEC filing delays, and whistle-blower charges. In 95
cases, the initial revelation date is identified in the Karpoff et al. (2008a, 2008b) database. These
events include announcements of an SEC informal inquiry or formal investigation, announcements
of a Wells Notice, the initiation of regulatory proceedings, class action lawsuits, and bankruptcy
announcements. Panel B reports on important announcements about the misrepresentation that
were made after the public revelation date. These subsequent events include announcements of an
informal SEC inquiry, formal SEC investigation, Wells Notice, initiation of regulatory proceedings,
initiation of class action lawsuits, and bankruptcy. There are a total of 844 such subsequent
announcements. Of these, 371 are follow-ups to the initial revelation date. Of these 371 cases, 274
have a third announcement, 147 have a fourth announcement, and 46 have a fifth announcement.
N
Mean (%)
Median (%)
t-Stat
−11.10
−15.00
−5.77
−11.70
−6.09
−1.03
−1.97
−3.73
−14.40
−19.90
−19.00
−8.55
−5.17
−4.62
N/A
−2.98
−5.12
−3.00
Panel A: Initial Public Revelation Date
All initial revelation dates
SEC-identified trigger event
Other initial revelation events
–SEC informal inquiry
–SEC formal investigation
–SEC Wells Notice
–Regulatory proceedings begin
–Class action lawsuits begin
–Bankruptcy
454
359
95
15
22
1
12
37
8
−18.20
−20.70
−8.90
−12.10
−9.32
−1.03
−6.29
−5.93
−20.40
Panel B: Important Subsequent Announcements
2nd announcement
3rd announcement
4th announcement
5th announcement
6th or higher announcement
All subsequent announcements combined
371
274
147
46
6
−9.61
−7.22
−3.52
−0.00
−13.76
−4.96
−3.97
−1.95
−0.90
−6.09
−12.41
−8.85
−4.88
0
−1.53
844
−7.28
−3.69
−15.30
for these 37 cases is −5.9%. Other less common revelation dates include the
announcement of a formal SEC investigation (22 events), an informal SEC
inquiry (15), the initial regulatory action and SEC release (12), and bankruptcy
filing (8). For all 95 of the revelation dates that are not identified by the SEC,
the mean 1-day market-adjusted return is −8.9%.
Averaging over all 454 initial revelation dates, the mean abnormal return
is −18.2% and the median is −11.1%. In the tests that follow we use data
from all 454 events. The results are qualitatively identical, however, if we
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1886
June 25, 2010
19:56
The Journal of FinanceR
limit the sample to the 359 events for which the SEC identified the trigger
date. Either way, these results indicate that public announcements that firms
violated financial reporting rules are associated with large declines in share
values. These are exactly the types of event that benefit short sellers.
Panel B of Table II reveals that share prices tend to decrease further when
additional news about the misrepresentation is revealed to the public. The announcements in this panel include SEC informal inquiries, SEC formal investigations, Wells Notices, the initiation of regulatory proceedings, the initiation of
class action lawsuits, and bankruptcies. A total of 371 of the 454 events have a
second announcement. The mean 1-day return for these 371-second announcements is −9.6%. A total of 274 events have a third announcement, with a mean
1-day return of −7.2%. Combining all 844 subsequent announcements in Panel
B, the mean 1-day return is −7.3% with a t-statistic of −15.3. These numbers
indicate that subsequent information about these firms’ financial misconduct—
even after the initial public revelation—also tends to be unfavorable.
B. Short Interest Data and Related Data
Our tests examine the ability of short sellers to depict misrepresentation
before it is publicly revealed, and thus we focus on short interest during the
violation period immediately before the initial public revelation dates that are
summarized in Panel A of Table II. Monthly short-interest data are obtained
from the New York Stock Exchange (NYSE), the American Stock Exchange
(Amex), and NASDAQ for the period January 1988 to December 2005.9 Short
interest reflects the open short positions of stocks with settlements on the last
business day on or before the 15th of each calendar month. Settlement, however,
takes a few days, and for a short sale transaction to be recorded in month t, it
must occur before or on the trade date. Before June 1995, the trade date was 5
days before the settlement date; currently, it is 3 days before. We define month
t as starting the day after the trade date of calendar month t−1 and ending
on the trade date of calendar month t. Raw short interest for firm i in month
t, SIit , is the percent of total shares outstanding in month t. The pooled mean
level of SIit over all months for all firms covered by the short interest data is
1.65%.
Monthly stock returns and market capitalization are constructed from daily
data obtained from CRSP using the month definition explained above. Some of
the analysis requires data on past returns and institutional ownership. Consequently, we use CRSP data from January 1987 through December 2005. We
obtain data on institutional ownership from the CDA/Spectrum database provided by Thomson Financial. The data, derived from institutional investors’
9 Daily data from January 1, 2005 through August 6, 2007 recently have become available to
researchers. These data, however, cover only a small number of the enforcement actions in our
sample. The daily data also do not contain information about short positions that are covered,
making it impossible to compute net changes in short interest. The monthly data therefore are
well suited to our tests.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1887
quarterly filings of SEC Form 13F, include quarterly holdings for each stock
for each quarter between December 1987 and December 2005.
C. Abnormal Short Interest
In addition to raw short interest, we examine three measures of abnormal
short interest. For firm i in month t, abnormal short interest equals
ABSI( j)it = SIit − E(SI( j)it ), j = 1, 2, 3,
(1)
where SIit is raw short interest and E(SI(j)it ) is the expected short interest
based upon one of three benchmarks j that reflect the firm’s characteristics.
The first benchmark, E(SI(1)it ), controls for the firm’s market capitalization,
book-to-market ratio, past stock performance, and industry. These controls
reflect findings by Dechow et al. (2001), Asquith et al. (2005), and Duarte et
al. (2006) that short interest is related to market capitalization, the book-tomarket ratio, and momentum. At the beginning of each month, each stock is
assigned to one of 27 portfolios constructed by independently sorting stocks
by size, book-to-market, and momentum, all measured at the end of the prior
month. Each of the 27 portfolios is further partitioned into industry groups
using two-digit SIC codes. We exclude the sample firms in constructing the
matching portfolios.
In particular, E(SI(1)it ) is the fitted value from a cross-sectional regression
that is estimated for each month t:
SIit =
medium
g=low
sgt Sizeigt +
medium
g=low
bgt BMigt +
medium
g=low
mgt Momigt +
K
k=1
φkt Indikt + uit .
(2)
The first three sets of explanatory variables are dummy variables that jointly
define the 27 size, book-to-market, and momentum based portfolios. For example, if firm i is assigned to the portfolio with the lowest market capitalization
in month t, then Sizei,low,t, = 1, Sizei,medium,t, = 0, and Sizei,high,t = 0. Industry
dummy Indit,k = 1 if firm i belongs to industry k in month t, and K is the total
K
number of industries. By construction, k=1
Indikt = 1 (so the intercept term
is omitted). Each monthly regression uses all firms listed on NYSE, Amex, or
NASDAQ that are not in our SEC enforcement action sample and for which
data on short interest, market capitalization, book-to-market, and momentum
are available over the period 1988 through 2005.
Table III reports the time-series averages of the coefficient estimates (excluding industry dummies) of the monthly cross-sectional regressions. The
associated t-statistics are computed with Newey–West (1987) corrections for
serial correlation using three lags. The base portfolio in this regression is the
portfolio with the highest market capitalization, book-to-market ratio, and momentum for each industry. This means that the coefficients are interpreted as
the difference between the short interest of the given portfolio and that of the
base portfolio. The results show that the largest firms have the highest short
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
The Journal of FinanceR
1888
Table III
Models Used to Calculate Abnormal Short Interest
For each month t, short interest (SI) is regressed on variables that are likely to explain the level
of short interest that is unrelated to short sellers’ information about financial misconduct. Short
interest (SI) is the number of shares shorted as a percentage of the number of shares outstanding.
The table reports the time-series means and t-statistics of the monthly coefficient estimates. For
Model 1:
SIit =
medium
sgt Sizeigt +
g=low
medium
bgt BMigt +
g=low
medium
g=low
mgt Momigt +
K
φkt Indikt + uit .
k=1
Explanatory variables include size, the book-to-market ratio, and momentum, all measured at the
beginning of month t. The explanatory variables are dummy variables. For example, if firm i is
assigned to the portfolio with the lowest market capitalization in month t, then Sizei,low,t = 1,
Sizei,medium,t = 0, and Sizei,high,t = 0. Model 2 includes dummy variables for share turnover and
institutional ownership, and Model 3 includes continuous variables for total accruals and insider
selling. All three regressions include industry dummies with Indikt = 1 if firm i belongs to industry
k in month t. Variable K is the total number of industries, and industry is defined using two-digit
SIC codes from CRSP. The sample includes all firms listed on NYSE, Amex, or NASDAQ that are
not in the SEC enforcement action sample and for which data are available during the 1988 to
2005 period. T-statistics are computed with Newey–West (1987) corrections for serial correlation
using three lags.
Sizelow
Sizemedium
BMlow
BMmedium
Momentumlow
Momentummedium
Model 1
(Used to
Calculate
ABSI(1))
Model 2
(Used to
Calculate
ABSI(2))
Model 3
(Used to
Calculate
ABSI(3))
−1.952
[−13.09]
−0.922
[−9.92]
0.345
[7.49]
−0.353
[−14.12]
0.402
[8.16]
−0.147
[−5.48]
−0.709
[−8.22]
−0.322
[−4.76]
0.270
[6.51]
−0.266
[−11.92]
0.454
[11.07]
0.093
[3.64]
−2.261
[−16.10]
−1.899
[−16.14]
−0.949
[−10.46]
−0.588
[−8.38]
−0.813
[−8.32]
−0.395
[−5.02]
0.264
[6.28]
−0.286
[−12.05]
0.466
[10.30]
0.093
[3.59]
−2.248
[15.73]
[−1.88]
[−15.72]
−0.931
[−8.94]
−0.531
[−6.84]
0.419
[7.38]
3.823
[10.28]
Yes
0.28
Turnoverlow
Turnovermedium
Institutional ownershiplow
Institutional ownershipmedium
Total accruals
Insider selling
Industry controls
Adj-R2
Yes
0.21
Yes
0.27
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1889
interest. Both the book-to-market ratio and momentum have U-shaped relations with short interest, as indicated by the different signs of b1 and b2 , and
m1 and m2 . The relation between the book-to-market ratio and short interest is
consistent with the finding in Dechow et al. (2001). The U-shaped relation between short interest and momentum also is documented by Duarte et al. (2006).
Stocks with the lowest book-to-market ratios and lowest past performance are
most highly shorted.
Our second measure of abnormal short interest, ABSI(2)it , includes additional controls for share turnover and institutional ownership, which D’Avolio
(2002) shows are related to short sales constraints. The coefficients reported
in the second column of Table III indicate that short interest increases with
both share turnover and institutional ownership. The fitted values from each
monthly cross-sectional regression are used to estimate E(SI(2)it ), the expected
amount of short interest for firm i in month t, which is in turn used to calculate
ABSI(2)it = SIit − E(SI(2)it ).
Our third measure of abnormal short interest, ABSI(3)it , expands the number
of control variables to include total firm accruals and insider selling. Healy
(1985), Dechow et al. (2010), and others show that accruals can be used to
manipulate earnings, and Agrawal and Cooper (2008) show that insider selling
is correlated with financial misconduct at many firms. Einhorn (2008) reports
that many short sellers base their positions on accruals and insider selling even
in the absence of any specific knowledge about the firm. ABSI(3)it reflects short
sellers’ information over and above their knowledge about accruals, insider
selling, or the other control variables.
Our measure of total accruals for firm i in month t is the same as that used
by Richardson et al. (2005):
Total accruals =
WCit + NC Oit + F I Nit
.
(Assetsit + Assetsi,t−12 )/2
(3)
Here, WCit is firm i’s change in noncash working capital. It is measured as
the change in current operating assets net of cash and short-term investments,
minus the change in current operating liabilities net of short-term debt. Noncurrent operating accruals, NCOit , is the change in noncurrent assets net of
long-term nonequity investments and advances, less the change in noncurrent
liabilities net of long-term debt. The change in net financial assets, FINit ,
is the change in short-term investments and long-term investments less the
change in short-term debt, long-term debt, and preferred stock. Total accruals
is measured using annual data, so it is the same for all months t in a given
fiscal year.
To measure Insider selling, we first calculate net insider selling in each month
as the difference between the shares sold and bought by insiders, divided by
the firms’ outstanding shares. Insider selling is then defined as the difference
between this measure of net insider selling and its mean over the previous 12
months. A higher number of this variable indicates a higher level of sales by
insiders.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1890
June 25, 2010
19:56
The Journal of FinanceR
The last column of Table III reports the means of the coefficients when Total
accruals and Insider selling are included in the monthly cross-sectional regressions for short interest. Short interest is positively related to both measures.
This indicates that short sellers respond to public information about accruals
and insiders’ trades, even in the absence of financial misrepresentation.
To check the robustness of our results, we replicated many of the tests reported in this paper using four alternate measures of abnormal short interest.
The first adds to ABSI(3) a control for the dispersion in analysts’ forecasts,
which Diether, Malloy, and Scherbina (2002) show is correlated with the cost
of selling short. The second adds the firm’s short interest at the beginning of
the violation period as a matching criterion in selecting comparison firms that
comprise the benchmark sample. The third uses each firm’s own level of short
interest before its violation period as its benchmark level of “normal” short
interest. And the fourth defines abnormal short interest as a binary variable
equal to one when raw monthly short interest exceeds the firm’s pre-violation
mean level by three standard deviations (this is similar to the method used
by Dyck, Morse, and Zingales (2010)). As reported in the Internet Appendix,
the results using any of these alternate measures are similar to those reported
here.
III. Do Short Sellers Identify Misrepresenting Firms?
A. Short Interest around the Revelation of Misrepresentation
Table IV (and Figure 2 below) report on monthly raw and abnormal short
interest during the 40-month period surrounding the month in which financial
misconduct is publicly revealed. We have at least some short interest data for
474 firms in the sample, but many of these firms do not have sufficient data
to calculate abnormal short interest in some months. This is for two reasons.
First, the data required to calculate abnormal short interest are not available
for all firms in all months. The data requirements are most severe for ABSI(3),
so our sample sizes typically decrease as we move from ABSI(1) to ABSI(3).
The second reason is that some firms enter the sample fewer than 19 months
before their public revelation dates, while some firms leave the sample or do
not have short interest data available in the months after public revelation.
In sensitivity tests, we find that the results do not change if we restrict the
sample to include only firms with data available for all months in the [−19, 0]
period.10
Average raw short interest SI t increases steadily from month −19 through
month 0, reaching a peak in month +5 before gradually decreasing through
month +20. The patterns for all three measures of abnormal short interest
are similar. The cross-sectional mean of ABSI(1)it, ABSI(1)t , is positive in
10 The results also are not sensitive to the [−19, +20] window. In general, abnormal short
interest becomes statistically significant around month −17, and it increases steadily, although
not monotonically, until 5 months after public revelation. We also find similar results when the
data are partitioned into 1988 to 1996 and 1997 to 2005 subperiods.
Table IV
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
−19
−18
−17
−16
−15
−14
−13
−12
−11
−10
−9
−8
−7
−6
−5
−4
−3
−2
−1
Month
1.916
2.254
2.401
2.508
2.463
2.603
2.736
2.943
3.178
3.212
3.362
3.419
3.362
3.424
3.432
3.242
3.422
3.564
3.743
SI
245
257
261
275
291
305
323
319
332
357
361
374
386
391
407
411
410
407
405
N
8.59
6.97
7.23
7.72
8.45
8.89
8.97
8.94
9.61
10.1
10.1
10.2
10.7
11.7
11.6
11.5
11.8
12.4
12.2
t-stat
0.320
0.577
0.695
0.821
0.797
0.911
1.067
1.260
1.476
1.512
1.743
1.746
1.667
1.646
1.719
1.492
1.632
1.760
1.890
ABSI(1)
212
219
226
235
245
260
273
267
277
294
309
323
336
340
349
361
361
364
361
N
Model 1
1.377
1.631
1.999
2.425
2.551
2.931
3.234
3.552
4.01
4.195
4.662
4.703
4.797
5.187
5.302
4.943
5.342
5.847
5.874
t-stat
0.137
0.380
0.515
0.644
0.593
0.703
0.823
1.062
1.270
1.261
1.443
1.387
1.355
1.214
1.296
1.132
1.247
1.308
1.451
ABSI(2)
212
219
226
235
245
260
273
267
277
294
309
323
336
340
349
361
361
364
361
N
Model 2
0.62
1.12
1.52
1.99
1.99
2.37
2.6
3.12
3.58
3.67
4.03
3.91
4.06
3.99
4.15
3.94
4.23
4.51
4.68
t-stat
0.218
0.528
0.605
0.734
0.682
0.857
0.915
1.190
1.374
1.382
1.425
1.399
1.417
1.412
1.513
1.302
1.441
1.449
1.651
ABSI(3)
179
190
196
207
217
224
239
234
241
256
262
278
290
293
301
310
314
315
314
N
0.88
1.37
1.57
2.03
2.04
2.51
2.59
3.19
3.51
3.63
3.52
3.57
3.8
4.06
4.28
4.02
4.38
4.48
4.75
t-stat
Model 3
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
(continued)
0.310
0.077
0.129
−0.053
0.175
0.058
0.276
0.184
0.008
0.042
−0.026
0.018
−0.005
0.101
−0.211
0.140
0.008
0.203
ABSI(3)
This table reports the mean levels of short interest (SI) and abnormal short interest (ABSI(j)) for firms in the financial misrepresentation sample
during the 40-month window around the revelation of financial misrepresentation. Month 0 is the month in which the financial misrepresentation was
publicly revealed. Abnormal short interest (ABSI(j)it ) for each event firm i in month t is the difference between the short interest and the predicted
short interest using the coefficients in month t using model j, j = 1,2,3. The time-series means of the coefficients from each model, j = 1,2,3, are
reported in Table III. Short interest (SIit ) is the number of shares shorted as a percentage of the number of shares outstanding in month t. N is the
number of firms used in calculating the average for each month in event time. The number of firms changes due to limited availability of data on
short interest or the variables used to calculate abnormal short interest. T-statistics test whether SI and ABSI(j) differ significantly from zero.
Short Interest and Abnormal Short Interest around the Revelation of Misconduct
jofi˙1597
19:56
1891
3.815
3.940
4.001
4.007
3.938
4.100
4.038
3.558
3.456
3.313
3.404
3.446
3.195
3.171
3.128
2.929
2.865
2.833
2.891
2.940
2.934
SI
374
324
308
300
288
286
277
276
265
258
256
250
251
242
243
242
238
235
232
227
222
N
12.3
11.7
11.8
11.9
11.4
11.1
10.6
11.3
12
12.1
12.2
12.1
11.7
11.3
11.1
10.7
10.8
11.4
11.3
11.3
10.9
t-stat
1.981
2.168
1.966
2.016
1.936
2.111
1.669
1.450
1.319
1.201
1.318
1.271
1.043
1.052
1.059
0.844
0.743
0.640
0.669
0.691
0.751
ABSI(1)
313
282
269
275
272
271
264
262
253
249
249
243
243
234
233
236
231
228
225
221
214
N
5.855
5.989
5.821
5.856
5.543
5.593
4.732
4.588
4.626
4.373
4.732
4.508
3.641
3.61
3.745
3.09
2.802
2.502
2.714
2.634
2.681
t-stat
1.513
1.379
1.395
1.489
1.502
1.638
1.276
1.099
0.949
0.779
0.849
0.925
0.677
0.667
0.724
0.615
0.527
0.337
0.486
0.496
0.498
ABSI(2)
313
282
269
275
272
271
264
262
253
249
249
243
243
234
233
236
231
228
225
221
214
N
Model 2
4.64
3.9
4.23
4.47
4.49
4.44
3.77
3.66
3.45
2.98
3.16
3.4
2.42
2.31
2.65
2.31
2.04
1.34
2.03
1.94
1.82
t-stat
1.682
1.756
1.676
1.736
1.730
1.892
1.451
1.263
1.059
0.897
1.035
1.046
0.782
0.762
0.839
0.747
0.704
0.407
0.557
0.616
0.620
ABSI(3)
274
241
232
233
233
231
230
228
222
219
218
213
211
205
203
206
202
201
199
196
190
N
4.62
4.33
4.45
4.57
4.56
4.52
3.85
3.78
3.55
3.19
3.49
3.53
2.51
2.4
2.78
2.54
2.48
1.48
2.13
2.22
2.09
t-stat
Model 3
0.030
0.074
−0.080
0.060
−0.006
0.162
−0.441
−0.189
−0.204
−0.162
0.138
0.012
−0.264
−0.020
0.077
−0.092
−0.043
−0.298
0.151
0.059
0.003
ABSI(3)
1892
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Month
Model 1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
Table IV—Continued
jofi˙1597
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
The Journal of FinanceR
19:56
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1893
4.5
4.0
Short interest or abnormal short interest
jofi˙1597
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
-19
-14
-9
SI
-4
ABSI(1)
1
ABSI(2)
6
ABSI(3)
11
16
Month relave to the revelaon of misconduct
Figure 2. Short interest around the revelation of misconduct. This figure plots the data
reported in Table IV, which contains the mean levels of raw and abnormal short interest in the
40-month window around the public revelation of misconduct for the sample of firms targeted
in SEC enforcement actions for financial misrepresentation from 1988 to 2005. Month 0 is the
month in which the misrepresentation was first publicly revealed. SI is the mean level of raw short
interest. ABSI(1), ABSI(2), and ABSI(3) refer to the three different measures of abnormal short
interest. Each measure of abnormal short interest equals raw short interest minus the predicted
short interest using the model parameters summarized in Table III.
month −19, indicating that these firms are more highly shorted than other
firms in the portfolio matched by size, book-to-market, momentum, and industry. However, ABSI(1)t does not differ significantly from zero at the 5%
level until month −17. In month −1, ABSI(1)−1 has a value of 1.890, meaning
that the misconduct firms’ short interest as a percentage of shares outstanding
is 1.890 percentage points higher than that of firms in the control portfolio.
Given that the unconditional mean short interest in any given firm-month is
only 1.65% of outstanding shares, this means that short interest in month −1
is more than double the unconditional sample mean level of short interest.
The second and third measures of abnormal short interest are smaller than
ABSI(1)t in every month, but both follow a similar pattern. In month −1,
ABSI(1)−2 equals 1.451 and ABSI(3)−1 equals 1.651. Thus, controlling for
share turnover and institutional ownership partly explains the abnormal increase in short interest that is reflected in ABSI(1)−1 . But controlling for total
accruals and insider selling does not further decrease the measure of abnormal
short interest. These results indicate that the build-up of short interest before
the public revelation of financial misrepresentation is not fully explained by
observable firm characteristics such as total accruals and insiders’ trades. As
reflected in Table III, short selling in general is sensitive to these characteristics. The build-up of short interest while firms misrepresent their financial
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1894
June 25, 2010
19:56
The Journal of FinanceR
statements, however, is attributable to something else. A plausible explanation is that short sellers act on private information or public information that
is not yet reflected in share prices. This is consistent with anecdotes (e.g.,
as in Einhorn (2008)) that short sellers identify overpriced shares through a
combination of fundamental analysis and private investigation.
The far right column in Table IV reports the monthly change in the third measure of abnormal short interest, ABSI(3)t . Fifteen of the 19 monthly changes
up through month 0 are positive, indicating that the build-up of short interest
is fairly steady. To measure the average monthly rate at which abnormal short
interest grows in the pre-revelation period, we estimate the following pooled
regression model using firm fixed effects:
ABSI( j)it = ai + δ·t+εit ,
(4)
where t is a time trend ranging from −19 to −1. We include firm fixed effects to
account for heterogeneity in short-selling activity across firms. Using the first
measure of abnormal short interest, ABSI(1)it , the estimate of the coefficient
for the time trend δ̂ is 0.073 with a t-statistic of 11.7. This indicates that
abnormal short interest increases by an average amount of 0.073 percentage
points in each of the 19 months leading up to the public revelation of financial
misrepresentation. The coefficient δ̂ using the second measure is 0.057 (t =
9.1), and for the third measure is 0.054 (t = 7.9).
The data from Table IV are illustrated in Figure 2. Raw short interest, as
well as all three measures of abnormal short interest, increase over the 19
months before the public revelation of financial misconduct, and slowly unwind
during the 20 months after public revelation. The Internet Appendix contains
an extension of Figure 2 that illustrates the overall pattern of short selling
around both the initiation and discovery of financial misconduct.
Notice that abnormal short interest does not immediately drop toward zero
in the months after the misconduct is publicly revealed. Instead, it remains
high for several months before gradually decreasing, remaining statistically
different from zero many months after the revelation. Our short interest data
do not provide the individual identities of each short seller, so it is possible
that the abnormal short interest after month 0 represents new short sellers
taking new positions in the stock. It is also possible that short sellers take
time to wind down their positions after month 0. Under either scenario, short
sellers can profit even after the initial revelation of misconduct. The evidence in
Panel B of Table II indicates that share prices tend to decline upon subsequent
announcements about the misconduct and penalty.
B. Short Interest and the Severity of Misrepresentation
The evidence in Table IV indicates that short sellers detect financial misrepresentation before it is publicly revealed. To probe this interpretation, we
examine whether the amount of short selling is related to the severity of the
misrepresentation. If short sellers are skilled at ferreting out information about
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1895
these firms’ overvaluation, we would expect short selling to be most pronounced
in the firms with the most severe reporting irregularities.
We examine cross-sectional differences in the amount of abnormal short
interest at month −1 using the following specification:
ABSI( j)i,−1 = λ0 + λ1 Severityi,−1 + λ2 Controlsi,−1 + εi, j = 1, 2, 3.
(5)
Here, ABSI(j)i,−1 is firm i’s abnormal short interest measured at the end
of month −1, and Severityi is a measure of the severity of the misconduct.
One potential measure of the misconduct’s severity is the drop in share value
when news of the misconduct is made public. Indeed, we find that the stock
return on the trigger date is significantly related to all three measures of abnormal short interest, indicating that short selling is particularly active before
public revelations of misconduct that precipitate large price drops. Unfortunately, this result does not directly tie short selling to the misrepresentation’s
severity. It is possible that short sellers have no specific knowledge of the misrepresentation, and are good only at anticipating large stock price declines.
It could also be argued that short sellers manipulate or engineer the large
stock price declines. To distinguish between these competing interpretations,
we need measures of severity that directly measure the extent of managers’
misconduct.
To do that, we use three different proxies for Severity: Fraud, Insider trading
charges, and Total accruals. The first of these proxies, Fraud, is a dummy
variable that equals one if the enforcement action includes fraud charges under:
(i) section 17(a) of the 1933 Securities Act, which covers fraudulent actions in
the sale of new securities; (ii) sections 10(a) or 10(b) of the 1934 Exchange Act,
which cover fraudulent actions in the purchase or sale of existing securities;
or (iii) sections 15 or 18 of the United States Code, which cover criminal fraud
charges. Contrary to popular use of the term “fraud,” specific charges of fraud
are not universal in SEC enforcement actions for misconduct. Fraud charges
are relatively difficult for the SEC or Department of Justice to prove, so they
tend to be included only when the financial misrepresentation is egregious and
costly (for a discussion, see Cox, Thomas, and Kiku (2003)). In our sample, 79%
of all enforcement actions include at least one fraud charge.
Our second proxy for severity, Insider trading charges, takes the value of one
if the enforcement action includes at least one charge of insider trading under
sections 10(b)5–1 and 10(b)5–2 of the Exchange Act. In our sample, 19% of all
actions include at least one such charge. Agrawal and Cooper (2008) conclude
that many managers trade on personal account when their firms’ books are in
error, and Karpoff, Lee, and Martin (2009) report that securities class action
settlements for financial misrepresentation are higher when insider trading
charges are included. These results indicate that insider trading charges are
associated with egregious and costly misrepresentations.
Our third measure of the severity of the misrepresentation is Total accruals,
as presented in equation (3). Healy (1985), Dechow et al. (2010), and others
have shown that accruals can be used to manipulate earnings. We hypothesize
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1896
June 25, 2010
19:56
The Journal of FinanceR
that the size of Total accruals correlates with the materiality of the financial
misrepresentation.
To verify that Fraud, Insider trading charges, and Total accruals are good
proxies for the severity of misconduct, we estimated the relation between each
proxy and the market-adjusted 1-day return on the revelation date, using the
same control variables as in Table V. The results, reported in the Internet
Appendix, indicate that each proxy is negatively and significantly related to the
abnormal return on the revelation date (p-values of 0.02 or smaller), indicating
that each is a good measure of severity. One way to think of these proxies is
that each is an instrumental variable for the stock price reaction to the public
revelation of misrepresentation. The use of these instrumental variables avoids
an errors-in-variables problem that would arise if we used the stock return to
measure severity in equation (5), because short interest and the stock return
upon public revelation both are simultaneously determined by the severity of
the misconduct.11
Models 1 to 3 in each panel of Table V report the results when equation (5)
is estimated using the three different measures of severity. Each variable is
measured at the end of the month just before the public revelation month. The
coefficients are all positive and statistically significant, implying that short
selling before the public revelation is significantly related to the severity of
the misrepresentation. In Model 1 of Panel A, the Fraud coefficient indicates
that short interest is 1.650 percentage points higher, on average, when fraud
charges are included in the enforcement action. Short interest is 2.034 percentage points higher when the enforcement action includes insider trading
charges. The Total accruals coefficient of 5.151 indicates that an increase from
the 10th percentile to the 90th percentile in the Total accruals measure corresponds to an increase in short interest of approximately 3.021 percentage
points. Similar estimates of the economic effects obtain from the results in
Panels B and C.
Model 4 in each of the panels includes all three Severity proxies in one
regression. In all three panels, the coefficients are positive, although only Total
accruals remains significant at the 5% level in all three panels. This could
reflect the high correlations among these proxies.
Among the control variables, high institutional ownership implies more
shares available on the stock loan market, which lowers the cost of shorting
and increases short interest. The Size coefficient is negative and significant,
indicating abnormal short interest is relatively small in large firms. This could
reflect the fact that size loads positively in constructing the benchmark levels
11 The Internet Appendix reports results for four alternate measures of misconduct severity: (i)
the monetary award in the private class action lawsuit related to the firm’s misconduct, (ii) the
monetary fines imposed by regulators, (iii) Karpoff et al. (2009) index of non-monetary penalties
imposed by regulators, and (iv) whether the firm subsequently declared bankruptcy. Measures (i),
(ii), and (iv) are significantly related to abnormal short interest, and yield results that are similar
to those for Fraud and Total accruals in Table V. When the alternate measures are included in
the tests, the coefficients on Fraud, Insider trading charges, and Total accruals all are positive,
generally with lower p-values than reported in the tables.
Table V
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
361
0.11
361
0.12
315
0.15
0.097
(0.00)
−0.827
(0.00)
0.141
(0.37)
0.005
(0.34)
1.872
(0.04)
5.151
(0.00)
315
0.16
0.094
(0.00)
−0.746
(0.00)
0.174
(0.27)
0.005
(0.42)
0.254
(0.83)
1.328
(0.13)
1.299
(0.12)
4.470
(0.00)
4
361
0.09
0.070
(0.00)
−0.582
(0.00)
0.136
(0.33)
0.006
(0.27)
0.258
(0.81)
1.847
(0.01)
1
361
0.09
0.074
(0.00)
−0.629
(0.00)
0.130
(0.35)
0.003
(0.51)
1.434
(0.08)
1.767
(0.01)
2
315
0.12
0.083
(0.00)
−0.740
(0.00)
0.135
(0.39)
0.006
(0.28)
1.573
(0.08)
4.650
(0.00)
3
315
0.13
0.079
(0.00)
−0.651
(0.00)
0.168
(0.28)
0.006
(0.31)
−0.189
(0.87)
1.547
(0.08)
1.046
(0.20)
4.050
(0.01)
4
314
0.10
0.077
(0.00)
−0.600
(0.00)
0.160
(0.31)
0.007
(0.22)
0.069
(0.95)
1.977
(0.02)
1
314
0.10
0.082
(0.00)
−0.682
(0.00)
0.158
(0.31)
0.004
(0.46)
1.428
(0.12)
1.860
(0.02)
2
314
0.11
0.082
(0.00)
−0.729
(0.00)
0.140
(0.37)
0.006
(0.30)
1.575
(0.08)
4.181
(0.00)
3
314
0.12
0.078
(0.00)
−0.641
(0.00)
0.172
(0.27)
0.005
(0.33)
−0.159
(0.89)
1.529
(0.08)
1.005
(0.22)
3.601
(0.01)
4
Short Sellers and Financial Misconduct
N
Adj-R2
0.089
(0.00)
−0.727
(0.00)
0.159
(0.26)
0.003
(0.60)
1.702
(0.04)
2.034
(0.01)
3
Panel C: ABSI(3)
June 25, 2010
Intercept
Momentum
Book-to-market ratio
Size
0.086
(0.00)
−0.692
(0.00)
0.158
(0.27)
0.005
(0.32)
0.798
(0.46)
1.650
(0.03)
2
Panel B: ABSI(2)
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
Control variables:
Inst. ownership
Total accruals
Insider trading charges
Severity measures:
Fraud
1
Panel A: ABSI(1)
The sample includes all SEC enforcement actions on NYSE/Amex/NASDAQ-listed firms for which data on short interest, market capitalization,
book-to-market, and momentum are available over the period 1988 to 2005. Fraud is a dummy variable that equals one if the enforcement action
includes fraud charges under Section 17(a) of the 1933 Securities Act or Section 10 of the 1934 Security Exchange Act. Insider trading charges is a
dummy variable that equals one if the action includes charges of insider trading. Total accruals is based on the measure in Richardson et al. (2005).
Institutional ownership is from the CDA/Spectrum database; Size is measured as the log of market capitalization; Book-to-market ratio is the ratio of
book assets to the sum of book liabilities and the market value of equity; and Momentum is calculated as the previous 12-month cumulative return.
ABSI( j)i,−1 = λ0 + λ1 Severityi,−1 + λ2 Controlsi,−1 + εi, j = 1, 2, 3.
This table reports the estimates and corresponding p-values for cross-sectional regressions that estimate the determinants of abnormal short interest
in the month immediately before the month in which financial misrepresentation is revealed to the public:
Determinants of Abnormal Short Interest at Month-1 Relative to Public Revelation
jofi˙1597
19:56
1897
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1898
June 25, 2010
19:56
The Journal of FinanceR
of expected short interest. Neither the book-to-market ratio nor momentum
has a significant impact on abnormal short interest.
The regressions in Table V analyze the determinants of short interest at
one point in the time line, namely, the last month before the exposure of the
misconduct. In the Internet Appendix, we report on how the change in abnormal
short interest from month −19 to month −1 is related to the severity of the
misconduct. The results from these tests indicate that short interest and its
build-up are positively related to the severity of the misrepresentation that
subsequently is revealed to the public. Short sellers not only pre-identify firms
that get into trouble for misrepresenting their financial statements, but they
also take larger positions when the misrepresentation is particularly egregious.
That is, short sellers appear to anticipate both the existence and severity of
financial misrepresentation.
C. Do Short Sellers Focus on Misrepresenting Firms?
The evidence in Tables IV and V (as well as the accompanying results in
the Internet Appendix) indicates that short sellers detect financial misrepresentation before it is publicly revealed, and that the extent of short selling is
sensitive to the severity of the misrepresentation. But these results are from
firms that, ex post, faced SEC sanctions for misconduct. They do not address
the question of whether short selling in general tends to predict the existence of
(yet undisclosed) misrepresentation. To explore this issue, we examine whether
high levels of abnormal short interest are related to the presence of financial
misconduct, using data from all firms for which we have short interest data.
Table VI reports results using ABSI(1), although the results are similar using
ABSI(2) or ABSI(3).
For each month t, we classify firms along two dimensions. Firms with ABSI(1)
in the top 5% are identified as “high short interest firms,” while those in the
bottom 95% are “low short interest firms.” If month t overlaps with an SECidentified violation period for firm i, it is designated as a “violation firm-month.”
This classifies every firm-month in the sample according to whether the firm
had high short interest and whether it was cooking its books.
Panel A of Table VI reports on the resulting 2 × 2 matrix. If short interest
tends to be high when firms misrepresent their financial statements, we should
see a higher-than-random concentration along the diagonal. That is, high short
interest firm-months should correspond with violation firm-months, and low
short interest firm-months should correspond with nonviolation firm-months.
This is exactly what we find. For example, 1.78% of all firm-months are in the
“violation” category. But among the high short interest firm-months, 4.18% are
in the violation category. A chi-squared test rejects the null hypothesis that
the short interest and violation categories are unrelated (χ 2 = 1,912, p-value =
0.00).
Panel A includes all firm-months for which we have short interest data. Panel
B excludes data from months between the public revelation of the misconduct
and the end of the SEC enforcement action. This affects only firms with SEC
Table VI
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
1,024,754
93.42
98.35
95.11
52,658
4.8
95.82
4.89
1,077,412
98.22
1,911.66
17,225
1.57
1.65
88.22
2,299
0.21
4.18
11.78
19,524
1.78
p-value
1,096,936
100
0
54,957
5.01
1,041,979
94.99
1,024,008
94.19
99.16
95.04
53,404
4.91
98.06
4.96
1,077,412
99.1
698.08
No Violation
8,719
0.8
0.84
89.19
1,057
0.1
1.94
10.81
9,776
0.9
p-value
Violation
1,087,188
100
0
54,461
5.01
1,032,727
94.99
Total
June 25, 2010
Chi-squared statistic
Total
Frequency
Percent
Row %
Column %
Frequency
Percent
Row %
Column %
Total
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
High
ABSI
Low
ABSI
Violation
Excluding Months after the Enforcement Actions
begins “High ABSI” = 1 if ABSI ≥ 95th Percentile
All Firm-Months
“High ABSI” = 1 if ABSI ≥ 95th Percentile
No Violation
Panel B:
Panel A:
Each panel groups all firm-months into four cells based on a two-way classification: (i) whether the amount of abnormal short interest is low or high,
and (ii) whether the firm subsequently is identified as having misrepresented its financial statements in that month. In Panel A, all firm-months
from the beginning of a firm’s violation to the end of its enforcement action are included in the “Violation” column. Panel B deletes all firm-months
between the public exposure of the violation to the end of the enforcement action. A firm-month is assigned to the “High ABSI” group if the firm’s
abnormal short interest in that month is above the 95th percentile of ABSI in the entire cross section of firms for that month. The table reports results
based on our first measure of abnormal short interest, ABSI(1), although similar results obtain for ABSI(2) and ABSI(3). The sample includes all
NYSE/Amex/NASDAQ stocks that are in the intersection of CRSP, Compustat, and the short interest data set.
Short Interest and the Presence or Absence of Financial Misconduct
jofi˙1597
19:56
Short Sellers and Financial Misconduct
1899
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1900
June 25, 2010
19:56
The Journal of FinanceR
enforcement actions, and has the effect of deleting observations for which short
interest may be high, even though the misconduct is public knowledge. Removing these observations yields results that are similar to Panel A. For example,
in Panel B, χ 2 = 698 with a p-value = 0.00. These results are consistent with
the hypothesis that there is a systematic relation between high short interest
and the presence of financial misconduct that has not yet been revealed to
the public. That is, short interest is a predictor of the existence of financial
misrepresentation in general.12
IV. Short Sellers’ External Effects on Other Investors
In this section, we examine whether short sellers confer external costs or
benefits on other investors. A potential external cost is that short selling may
exacerbate a downward price spiral when the misconduct is publicly revealed.
A potential benefit is that short selling may help uncover the misconduct.
A second potential benefit is that short selling may dampen the stock price
inflation that occurs when the firm’s books are in error.
A. Short Selling and the Share Price Reaction to News of Misconduct
Critics of short selling argue that it can cause prices to deviate from fundamental values, particularly when bad news hits the market. In written testimony for the U.S. House Committee on Financial Services, for example, MBIA
Inc. argues that short sellers increased the downward price pressure on insurers that faced large losses in the U.S. mortgage markets (see Wilchins (2008)).
Short selling, according to this view, creates a cascade of selling that leads to
overreaction to bad news and drives prices down too far. The SEC’s October
2008 moratorium on naked short selling in selected financial institutions was
based in part on this theory, as SEC Chairman Christopher Cox argued that
short selling contributed to large share price declines at such firms as Lehman
Brothers, Bear Stearns, Fannie Mae, and Freddie Mac (e.g., see Cox (2008), Q1
Zarroli (2008)).
To investigate this argument, we examine how short selling is related to the
stock price reaction when financial misconduct is revealed to the public. In
particular, we estimate the following cross-sectional regression:
ARi = a + f1 ABSI( j)i,−1 + f2 Severityi + f3 Controlsi + ei , j = 1, 2, 3, (6)
where ARi is the market-adjusted return on the day misrepresentation is first
publicly revealed, and ABSI(j)i,−1 is firm i’s abnormal short interest in the
month before the month of the revelation date. If the announcement day abnormal return is sensitive to the severity of the misconduct, f 2 should be negative.
12 For a more comprehensive investigation of variables that predict financial misconduct, see
Dechow et al. (2007). Our results indicate that abnormal short interest should be included in such
tests. The Internet Appendix reports that we obtain similar results when “high short interest” is
defined as the top 10% of ABSI(1).
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1901
Table VII
Short Sellers and the Market Penalty for Misrepresentation
This table reports the estimates and corresponding p-values for cross-sectional regressions that
estimate the determinants of the market-adjusted abnormal return on the day financial misrepresentation is publicly revealed (ARi ):
ARi = a + f1 ABSI(1)i,−1 + f2 Severityi + f3 Controlsi + ei .
The sample includes all SEC enforcement actions for NYSE/Amex/NASDAQ-listed firms for which
data are available over the period 1988 to 2005. Variables are defined as in Tables IV and V.
Model
Abnormal short interest, ABSI(1)i,−1
1
2
3
4
5
−0.314
(0.10)
−0.231
(0.23)
−9.695
(0.00)
−0.185
(0.33)
−0.191
(0.34)
−0.052
(0.79)
−7.475
(0.01)
−9.050
(0.00)
−11.181
(0.02)
−0.052
(0.30)
1.059
(0.08)
0.025
(0.35)
−0.009
(0.17)
−12.31
(0.00)
295
0.12
Fraud
−11.774
(0.00)
Insider trading charges
−0.073
(0.15)
1.699
(0.00)
0.047
(0.10)
−0.012
(0.07)
−24.04
(0.00)
−0.056
(0.26)
1.367
(0.01)
0.042
(0.13)
−0.012
(0.06)
−15.42
(0.00)
−0.076
(0.12)
1.627
(0.00)
0.036
(0.18)
−0.007
(0.27)
−21.12
(0.00)
−14.850
(0.00)
−0.064
(0.22)
1.423
(0.02)
0.035
(0.21)
−0.012
(0.07)
−21.10
(0.00)
340
0.04
340
0.07
340
0.10
295
0.05
Total accruals
Institutional ownership
Size
Book-to-market ratio
Momentum
Intercept
N
Adj-R2
If, in addition, short selling causes an overreaction that is not related to the
severity of the misconduct, f 1 should be negative as well.
Table VII presents the results using our first measure of abnormal short
interest, ABSI(1). In Model 1, the coefficient for ABSI(1)i,−1 is negative and
significant at the 10% level. This appears to provide support for the view
that short selling exacerbates the price drop when bad news hits the market. But Model 1 does not control for the severity of the misconduct. The
negative relation between the abnormal stock return and abnormal short interest might simply reflect short sellers’ tendencies—documented earlier in
Table V—to take larger positions when the misrepresentation is particularly
bad.
When we include our measures of severity, as in Models 2–5, the coefficient
for ABSI(1)i,−1 becomes statistically insignificant. Instead, the coefficients for
the severity measures are negative and significant. The results are more pronounced when we use ABSI(2)i,−1 or ABSI(3)i,−1 to measure abnormal short
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1902
June 25, 2010
19:56
The Journal of FinanceR
interest; in these cases the abnormal return is not significantly related to abnormal short interest, but is strongly and negatively related to all three measures of misconduct severity. These results indicate that short-selling activity
does not have a significant, independent effect on the market reaction to news
of financial misconduct. This is inconsistent with the view that short selling
causes an unwarranted downward spiral in the stock price when bad news is
announced.
We also examine the stock price behavior after the initial disclosure of misconduct. If short selling causes overreaction—either in the short term or long
term—then we should see differences in the stock price paths of our sample
firms that correspond to differences in short interest. The results, however, do
not support an overreaction story. In results reported in the Internet Appendix,
we partitioned the sample according to abnormal short interest in month −1.
We find no evidence of overreaction on the public revelation day (day 0) in either the high-ABSI or low-ABSI groups, as there is no price reversal after day 0
in either group. We also find no significant difference in the cumulative excess
returns between the high-ABSI and low-ABSI groups over horizons up to 180
days after day 0. These results indicate that the market responds to the severity of misconduct. But there is no evidence that short selling in and of itself
imposes additional downward price pressure, either on the public revelation
day or afterwards.
B. Do Short Sellers Help Expose Financial Misrepresentation?
B.1. Short Interest and the Time to Revelation
Short selling advocates (e.g., Einhorn (2008)) argue that short sellers generate external benefits by helping to expose financial misrepresentation. To
our knowledge, this assertion has not been tested, although a recent paper
by Dyck et al. (2010) provides some evidence. Dyck et al. examine spikes in
short interest before the announcements of security class action lawsuits from
1996 to 2004, most of which are for financial misrepresentation. An event
is labeled “detected by short sellers” if the raw short interest during the
3 months before the filing date exceeds the firm’s short interest in the prior
year by three standard deviations. Dyck et al. conclude that between 3.4%
and 14.5% of their 216 events are detected by short sellers. This estimate
suggests that short sellers play a modest role in helping to uncover financial
misconduct.
We examine this issue by estimating survival models that measure how short
selling affects the time it takes for misrepresentation to be publicly revealed.
Specifically, we model the logarithm of the time to revelation, log(Mi ), as
log(Mi ) = β Xi + εi .
(7)
Here, Mi is the number of months from the beginning of firm i’s violation until
its revelation, Xi is the vector of possibly time-varying covariates assumed
to influence the time until public revelation, and β is a vector of regression
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1903
parameters that we estimate. The error term εi is assumed to follow the logistic
distribution.13
In estimating the model, we use data from all months from violation until
the revelation of misconduct. Letting Tit denote the number of months from
the start of violation, we use all firm-months such that 0 < Tit ≤ Mi . The explanatory variables Xi are measured at the beginning of each month t. For each
month t, we observe the following vector [t, Revelationi , Xi ], where Revelationi is
a dummy variable that equals one if the firm’s misconduct is revealed in month
t (Mi = Tit ) and zero otherwise (Mi > Tit ). A log-likelihood function can then
be constructed to estimate the parameter vector β. To mitigate contamination
from outliers, only violations that last between 1 and 10 years are included
in the estimates we report, although the results are not sensitive to the inclusion of all events. In the data matrix Xi , we include controls for institutional
ownership, size, book-to-market, and momentum.
Model 1 in Panel A of Table VIII reports the impact of abnormal short interest on the time to exposure. We report results using ABSI(1), although the
results are similar using ABSI(2) and ABSI(3). The coefficient is −0.028 and is
statistically significant, indicating that short selling is associated with a more
rapid exposure of the misconduct. In Models 2 to 5, we include controls for the
severity of the misconduct. The coefficients for Fraud in Model 2 and for Total
accruals in Model 4 both are negative and statistically significant, consistent
with the view that severe misrepresentations are discovered relatively quickly.
These results continue to hold in Model 5, which includes all three severity
measures together. Most importantly for our investigation, the coefficient for
abnormal short interest is negative and significant in all model specifications.
This implies that short selling is positively related to the speed with which
financial misrepresentation is uncovered.
The coefficient estimates from the parametric log-logistic model allow us to
quantify how abnormal short interest affects the time until misrepresentation
is discovered. Consider short interest in firms whose violations have been ongoing for 12 months. Using the Model 5 estimates and inserting mean values
for all other variables, a firm at the 75th percentile of abnormal short interest
eventually will have its misconduct uncovered 8 months sooner than a firm at
the 25th percentile. As reported in Table I, the median time-to-revelation is 26
months, so an 8-month reduction in the time to revelation represents a significant impact of short selling on the time it takes to uncover the misconduct.
B.2. Endogeneity
Table V shows that short sellers take particularly large positions when
the misrepresentation is particularly egregious. Suppose that egregious
13 We obtain qualitatively identical results using a discrete version of Cox’s (1972) semiparametric proportional hazard model. The Cox model requires no assumption about the distribution
of ε i in equation (7). An advantage of the parametric specification, however, is that it enables us to
obtain quantitative estimates of the impact of short selling on the time to revelation, as reported
below.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
The Journal of FinanceR
1904
misrepresentations are discovered quickly, not due to short selling, but rather
because their severity prompts a fast response from investors or regulators.
In this case, we would not conclude that short interest accelerates the time to
discovery, but rather that both short interest and time-to-discovery are driven
by the severity of the misconduct.
This concern is mitigated somewhat by including measures of misconduct
severity in the empirical tests, as in Panel A of Table VIII. The severity measures, however, may be imperfect. To further control for a possible omitted
Table VIII
Short Selling and the Public Exposure of Financial
Misrepresentation
This table reports the coefficients estimates and corresponding p-values for the following parametric survival model:
log(Mi ) = β Xi + εi .
Mi is the month in which firm i’s misconduct is revealed to the public. The regression is estimated
using data from all months in the violation period through the month of public revelation. Xi includes variables that are likely to affect the exposure of misconduct, most importantly, abnormal
short interest. The error term is assumed to follow a logistic distribution. The sample includes
all misrepresentations for which data are available over the period 1988 to 2005. Panel A reports
the results using the first measure of abnormal short interest, ABSI(1). Results are similar using
the alternate measures of abnormal short interest, ABSI(2) and ABSI(3). Panel B reports results
using instrumental variables for each of the three different measures of abnormal short interest.
Variables are defined as in Tables IV and V.
Panel A: Direct Tests (Using ABSI(1) to Measure Abnormal Short Interest)
Model
Abnormal short interest (ABSI(1))
1
2
3
4
5
−0.028
(< 0.001)
−0.025
(< 0.001)
−0.323
(< 0.001)
−0.028
(< 0.001)
−0.026
(< 0.001)
−0.023
(< 0.001)
−0.480
(< 0.001)
0.122
(0.13)
−0.197
(0.07)
0.001
(0.44)
−0.004
(0.83)
0.002
(0.003)
0.001
(0.04)
5.263
7,160
Fraud
−0.008
(0.91)
Insider trading charges
−0.0003
(0.85)
0.023
(0.16)
0.002
(0.001)
0.001
(0.05)
−0.0001
(0.92)
0.012
(0.47)
0.002
(0.002)
0.001
(0.03)
−0.0003
(0.85)
0.023
(0.16)
0.002
(0.001)
0.001
(0.05)
−0.228
(0.05)
−0.0001
(0.96)
0.024
(0.19)
0.002
(0.002)
0.001
(0.05)
4.945
8,902
5.193
8,902
4.947
8,902
4.899
7,160
Total accruals
Institutional ownership
Size
Book-to-market ratio
Momentum
Intercept
N
(continued)
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1905
Table VIII—Continued
Panel B: Instrumental Variable Tests
Abnormal Short Interest
Instrumental variable for ABSI(j)
Fraud
Insider trading charges
Total accruals
Institutional ownership
Size
Book-to-market ratio
Momentum
Intercept
N
ABSI(1)
ABSI(2)
ABSI(3)
−0.261
(0.00)
−0.485
(0.00)
0.099
(0.20)
−0.490
(0.00)
0.0020
(0.11)
0.052
(0.00)
−0.0068
(0.57)
0.0015
(0.00)
−0.206
(0.00)
−0.594
(0.00)
0.163
(0.05)
−0.524
(0.00)
0.0010
(0.46)
0.032
(0.07)
−0.0106
(0.42)
0.0019
(0.00)
−0.158
(0.00)
−0.540
(0.00)
0.154
(0.08)
−0.592
(0.00)
0.0008
(0.59)
0.038
(0.04)
−0.0096
(0.48)
0.0017
(0.00)
4.732
4,922
4.986
4,922
4.974
4,922
variables bias in estimating equation (7), we construct an instrumental variable for short interest. The instrumental variable is the fitted value from the
following cross-sectional model, which is estimated for each month t:
ABSI( j)it = δ0t + δ1t Optionsit + εit ,
j = 1, 2, 3.
(8)
In equation (8), Optionsit is a dummy variable set equal to one if the stock has
listed options on the CBOE in month t. Diether, Malloy, and Scherbina (2009) Q2
find that short selling activity is positively related to the availability of options
markets trading. This is for two reasons. First, listed options can decrease the
cost of hedging short positions. And second, firms with listed options may be less
expensive to borrow and sell short since stocks with options tend to be larger
and more liquid. Consistent with such prior findings, we find that the mean
of the cross-sectional coefficients δ̂1t in equation (8) is positive and significant
using all three measures of abnormal short interest. Using coefficient estimates
δ̂0t and δ̂1t from equation (8), we create an instrumental variable equal to the
j) .14
fitted value ABSI(
i,−1
14 In most tests, ABSI(
j)
i,−1 satisfies the exclusion restriction for an instrumental variable, as
it is not significantly correlated with the residual in the time-to-revelation models. An exception
is Model 2 in Panel B of Table VIII, for which the instrument is significantly correlated with the
residual. For discussions of instrumental variable tests with hazard models, see Abbring and Van
den Berg (2005) and Bijwaard (2008).
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1906
June 25, 2010
19:56
The Journal of FinanceR
Panel B of Table VIII reports the results of the instrumental variable estimation. We use the same specification as in Model 5 of Panel A, except that
j) . Using any
we replace ABSI(j)i,−1 with the instrumental variable ABSI(
i,−1
of the three measures, the coefficient for the instrumental variable is negative
and statistically significant (p < 0.001). This relation is robust to controls for
the severity of the misrepresentation, which also are positively related to the
speed with which misrepresentation is uncovered.
The evidence from both Panels A and B in Table VIII indicates that short
sellers play an important role in helping to uncover financial misrepresentation. These results are contrary to the estimates by Dyck et al. (2010), which
imply that short sellers play only a modest role in helping to unveil business
misconduct.
C. External Effects of Short Selling on Prices and Share Quantity
In addition to impacting the time to revelation, short selling can affect share
prices and the number of shares held by other investors. Uninformed investors
can benefit if short selling keeps prices closer to their full-information values.
But they also can be harmed by the extent of short sellers’ profits.
Figure 3 illustrates these two external effects. In the figure, Pactual represents
the price at which shares trade in any given month t during the violation
period. In the absence of informed short selling, the inflated price presumably
would have been higher, Phigh . The difference, Phigh – Pactual , represents short
sellers’ price impact. When uninformed investors trade with each other, the net
distributional effect of the price impact is zero. But when uninformed investors
buy from informed sellers, the price impact conveys an external benefit, which
is illustrated by Area B (for “Benefit”) in Figure 3. We consider trades with
two types of informed sellers: insiders and the firm itself. Area B equals the
price impact, Phigh – Pactual , times the net number of shares sold by insiders or
issued by the firm in month t. It is a measure of uninformed investors’ savings
when they buy from informed parties—insiders or the firm itself—who are most
likely to gain from the artificial inflation in share prices that occurs when the
firm issues falsified financial statements.
Short sellers also impose external costs on uninformed investors. Borrowing
shares to sell short increases the number of shares held by uninformed investors.15 Uninformed investors buy these shares during the violation period,
when the shares typically are overpriced. In Figure 3, Ptrue is the value that
would obtain if the firm’s financial statements were not in error. Every short
sale occurs at a price that is inflated by the amount Pactual – Ptrue . The total external cost to uninformed investors, represented by Area C (for “Cost”), equals
Pactual – Ptrue times the number of shares sold short.
15 For a discussion, see Apfel et al. (2001). Apfel et al. also point out that, because it increases
the number of shares held long, short selling makes it difficult to identify which shareholders have
standing to sue in 10b-5 class action lawsuits for financial fraud. This is a potential cost of short
selling that our tests do not measure.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1907
Hypothetical price without abnormal short selling (Phigh)
Area B:
Benefits to
uninformed
investors
Actual price with short selling (Pactual)
Area C:
Costs to
uninformed
investors
Shares sold by the
firm and insiders
Hypothetical price with no misconduct (Ptrue)
Shares sold by the
firm and insiders
+ New shares
created by short
sellers
? #shares outstanding
Figure 3. External effects of short sellers on uninformed investors. This figure illustrates
two external effects of short sellers on uninformed investors during a period in which the firm’s
financial statements are in error. The top line (Phigh ) represents the hypothetical price at which
shares would trade in a given month t if there was no abnormal short selling. It is calculated using
a model of monthly share returns with an additional regressor that estimates the marginal impact
on monthly returns from abnormal short interest. The middle line (Pactual ) represents the observed
price in the month. The bottom line (Ptrue ) represents the hypothetical price at which shares would
trade if there was no financial misconduct (or abnormal short selling). Shares sold by the firm and
insiders is the net change in the number of outstanding shares in month t plus the net number of
shares sold by insiders in month t. New shares created by short sellers is the net change in abnormal
short interest. Area B represents the external benefits to uninformed investors because abnormal
short interest dampens the price inflation during the misconduct period. Area C represents the
external cost to uninformed investors due to the additional shares created by short selling, and
equals short sellers’ profit.
Notice that Area C represents short sellers’ gain. If Area C equals Area B,
short sellers internalize their external benefits exactly. If B is positive but less
than C, then short sellers generate external gains for uninformed investors,
but these gains are more than offset by their profits. If B is greater than C,
then short sellers generate external gains that exceed their profits.
To estimate Areas B and C, we need estimates of Phigh – Pactual and Pactual –
Ptrue . For Phigh , we first estimate a cross-sectional model for share returns, retit ,
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1908
June 25, 2010
19:56
The Journal of FinanceR
using firms not in the SEC enforcement action sample:
retit = β0 + β1 Sizei,t−1 + β2 BTMi,t−1 + β3 Momi,t−1 +
K
Indik,t−1
k=1
+ β4 ABSI( j)i,t−1 + εi ,
(9)
Equation (9) is estimated for each month t, and the explanatory variables
are measured at the end of the prior month, t – 1. The key to this model is
the inclusion of abnormal short interest, ABSI(j)i,t−1 . The mean estimate β̂4 is
negative, indicating that short selling does indeed have an impact on prices.
Using ABSI(1), for example, the mean of the monthly coefficients β̂4 is −0.091
with a t-statistic of −10.52.
high
The hypothetical return if abnormal short interest were zero, retit , is
high
retit
= rit − β̂4 ABSI( j)i,t−1 .
(10)
For each firm-month, we calculate the hypothetical cumulative return,
high
cumretit , from the beginning of the violation:
high
cumretit
=
t
high
retiτ .
(11)
τ =1
high
The hypothetical stock price, Phigh , is then calculated as P0 ∗ cumretit , where
P0 is the stock price at the beginning of the violation period. In measuring P0 ,
we adjust for stock splits using the cumulative adjustment factor provided
by CRSP. The difference between Phigh and the actual price at the end of the
contemporaneous month is given by Phigh – Pactual .
We use two estimates of Pactual – Ptrue . Our upper bound estimate equals
the actual price in month t minus the price per share immediately after the
misconduct is revealed. Our lower bound estimate equals 24.53% of this difference. The rationale for the upper bound estimate is that the postrevelation
price reflects investors’ valuation after they adjust for the news that the price
previously had been inflated by falsified financial statements. The rationale for
the lower bound estimate is that, as Karpoff et al. (2008b) report, the postrevelation share price falls more than the price inflation that we can attribute to
the misconduct. The price drop also reflects investors’ expectations of future
legal penalties and the firm’s reputation loss. Karpoff et al. (2008b, page 600)
estimate that 24.53% of the price drop represents the amount by which prices
were inflated by the financial misrepresentation.
We use these variables to calculate Areas B and C for each month of the
violation period as a percentage of the firm’s market capitalization. For each
firm, we sum the monthly estimates of Area B to obtain a firm-specific estimate
of short sellers’ external benefits. Similarly, we sum the monthly estimates of
Area C to obtain a firm-specific estimate of short sellers’ profits, which also is
a measure of the external costs imposed on uninformed investors.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Short Sellers and Financial Misconduct
1909
Table IX
Short Sellers’ External Effects on Uninformed Investors
This table reports estimates of the external benefits and costs for uninformed investors from short
sellers’ trading during the period that the firms’ books were in error, using ABSI(1) to measure
abnormal short interest (N = 359).%Shares sold by the firm and insiders is the net change in
shares outstanding plus net insider sales, expressed as a percentage of shares outstanding at the
beginning of the month, and cumulated over all months of the violation period. Short sellers’ price
impact, Phigh – Pactual , is the difference between the hypothetical price in the absence of abnormal
short interest and the actual month-end price, expressed as a percentage of the actual share price
at the beginning of the month. External benefit is the sum of the monthly estimates of Area B in
Figure 3. Each monthly estimate equals the product of %Shares sold by the firm and insiders and
Short sellers’ price impact, and is expressed as a percentage of the firm’s equity value. %New shares
created by short sellers is the increase in ABSI(1) from the prior month, expressed as a percentage
of shares outstanding at the beginning of the month, and cumulated over all months of the violation
period. Short sellers’ profit per share, Pactual – Ptrue, is the difference between the actual price and
the price when news of the misconduct is first revealed to the public, expressed as a percentage
of the actual share price at the beginning of the month. External cost is the sum of the monthly
estimates of Area C in Figure 3. It equals the product of %New shares created by short sellers and
Short sellers’ profit per share. Net external effect is the difference between External benefit and
External cost. Each variable is measured in each month of a firm’s violation period, and summed
over all violation period months. The table reports the mean and median of the cross section of
firm-specific measures and, in brackets, the 95% confidence interval for the mean. T-statistics and
confidence intervals are computed from the cross section of firm-specific measures.
Mean
[95% Conf. Interval]
External benefit
% Shares sold by the firm and insiders
Short sellers’ price impact, Phigh – Pactual (% of share price)
External benefit (sum of monthly estimates of Area B)
External cost (=short sellers’ profits)
% New shares created by short sellers
Short sellers’ profit per share, Pactual – Ptrue (% of share price)
External cost (sum of monthly estimates of Area C)
Net external effect (sum of monthly Area B – Area C)
Main estimate
Using a lower-bound estimate of external cost
45.65
[26.24, 65.06]
2.41
[1.61, 3.21]
1.67
[0.25, 3.09]
1.12
[0.54, 1.70]
12.13
[2.35, 21.91]
0.58
[−0.05, 1.21]
1.09
[−0.80, 2.98]
1.53
[0.01, 3.05]
Median
8.34
0.22
0.00
0.08
30.44
0.08
–0.06
0.00
Table IX reports on the resulting estimates using ABSI(1) to measure abnormal short interest. Estimates using ABSI(2) and ABSI(3) are in the Internet
Appendix. During the violation period, insiders and the firm jointly sell shares
that average 45.65% of the firm’s outstanding common stock, with a 95% confidence interval of [26.24%, 65.06%] (the t-statistic is 4.61). The time-series
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1910
June 25, 2010
19:56
The Journal of FinanceR
mean of short sellers’ monthly price impact, expressed as a percentage of the
beginning-of-month share price, is 2.41%. Together, these imply that short
sellers’ external benefit to uninformed investors—the sum of the monthly measures of Area B—is 1.67% of the firm’s market capitalization (95% confidence
interval of [0.25%, 3.09%]).
Note that the median values are much smaller. The quantity distribution
is skewed right, as a small number of firms issue a large number of shares.
The distribution of price impact also is skewed right, and for the median firm,
short sellers’ external benefit is negligible. Thus, short sellers generate external benefits, on average. But these benefits are negligible for the median firm.
Further examination reveals that the right skew in the distribution of external
benefits reflects a small number of firms that issued stock during their violation periods, including Cendant Corp., Waste Management Inc., Triton Energy
Corp. America Online, Inc., and Royal Ahold NV.16
Panel A also reports on short sellers’ profits. During the violation period, the
average change in ABSI(1) is 1.12% of outstanding shares (confidence interval
of [0.54%, 1.70%]), and our upper bound mean estimate of Pactual –Ptrue is 12.13%
of share value per month. Averaging across firms, short sellers’ mean profit is
0.58% of the firm’s market capitalization. That is, short sellers cumulatively
generate profits on the positions they take during the violation period that
average 0.58% of the firm’s equity value.
The mean net effect, equal to the sum of the differences between each firm’s
monthly estimates of Areas B and C, is 1.09% of equity value. Note, however,
that the confidence interval [−0.80, 2.98] indicates that this estimate is noisy.
Furthermore, the median value is negative, albeit small in magnitude (−0.06%
of equity value). This indicates that, for most firms, short sellers internalize
their external benefits via their trading profits, generating no net external
benefits for shareholders. But in a small number of firms, short sellers’ net
external benefits are positive and large. The net external benefits in these
relatively few cases are large enough to generate a positive mean estimate of
the external benefit.
If we use the lower bound estimate of short sellers’ external costs, the net
external benefit is larger. The lower bound estimate of external costs is 0.14% of
equity value, increasing the mean estimate of short sellers’ net external benefits
to uninformed shareholders to 1.53% of equity value (confidence interval of
[0.01, 3.05]). This larger estimate of net benefits is appropriate if we think of
Ptrue as the price that would have obtained if the firm had never misrepresented
its financials in the first place, that is, if we exclude from the definition of Ptrue
any legal penalties and reputation losses that accrue to firms that misrepresent
their financials.
These results indicate that short sellers generate external benefits for uninformed investors, which in the median case they internalize with their trading
16 This finding is consistent with arguments that one reason firms misrepresent their financials
is to issue new shares at a favorable price (e.g., see Dechow et al. (1996), Efendi, Srivastava, and
Swanson (2007)).
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1911
profits. For a small number of firms, however, the external benefits are large
enough to affect the overall mean estimates. Short sellers generate net external
benefits particularly when they take positions in misrepresenting firms that
issue new (overpriced) shares to uninformed investors.
V. Conclusions
Short sellers attract a lot of attention. They are blamed for manipulating
and depressing share values, and for exacerbating price declines when bad
news is announced. They also are credited with improving financial markets’
informational efficiency. We provide evidence on one aspect of short sellers’
effects on markets by examining short selling before the public revelation that
firms misrepresented their financial statements. Such revelations are material events, as they are associated with an average 1-day share price decline
of 18%. Short sellers anticipate such announcements, as abnormal short interest builds steadily in these stocks during the 19-month period before the
public revelation. The amount of short selling increases with the severity of
the misrepresentation, indicating that short sellers are sensitive to the characteristics of the misconduct. High short interest also concentrates in firms that
misrepresent their financials, compared to firms that do not.
These results imply that short sellers are proficient at identifying financial
misrepresentation before the general investing public. Our measures of abnormal short interest control for firm characteristics that are known to correlate
with (and possibly motivate or facilitate) short selling, including firm size, the
book-to-market ratio, momentum, share turnover, institutional ownership, insider trading, and total accruals. But short sellers do not merely track such
firm characteristics. Rather, their positions appear also to be based on private
information or a superior synthesis of public information about whether firms
are cooking their books.
We also examine short sellers’ external effects on other investors. Contrary
to some claims (e.g., see Wilchins (2008) Zarroli (2008)), short selling does not
exacerbate the decline in share prices when bad news is announced. To the contrary, short selling conveys external benefits to uninformed investors, in two
ways. First, short selling increases the speed with which financial misrepresentation is detected. Among firms that have been misrepresenting their financials
for 12 months, our point estimates indicate that a firm at the 75th percentile
of abnormal short interest will be publicly revealed 8 months sooner than a
firm at the 25th percentile. Thus, short selling not only anticipates financial
misconduct but it also helps expose the misconduct.
The second external benefit is that short selling mitigates the mispricing that
occurs when firms misrepresent their financial statements. This price impact
conveys offsetting benefits and costs to uninformed investors who trade with
each other. But for uninformed investors who purchase shares from insiders or
the firm, the benefits can be substantial. We estimate that this benefit equals
1.67% of the firm’s market capitalization. Short sellers internalize some of this
benefit, profiting by 0.58% of the firm’s equity value, on average. Net of this
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
1912
June 25, 2010
19:56
The Journal of FinanceR
profit, short sellers’ net external benefit is still positive, averaging 1.09% of the
firm’s equity value.
These results indicate that short sellers tend to ferret out and help uncover
financial misconduct by corporate managers. Short sellers profit from their positions in firms whose misconduct subsequently is revealed. Yet, even net of
these profits, short sellers generate external benefits for uninformed investors.
By improving market efficiency through its effects on prices, short selling offsets some of the harm imposed on uninformed investors who unwittingly buy
shares from firms and insiders while the firm’s books are in error.
REFERENCES
Abbring, Jaap H., and Gerard J. Van den Berg, 2005, Social experiments and instrumental variables with duration outcomes, Tinbergen Institute Discussion paper no. TI 05-047/3.
Agrawal, Anup, and Tommy Cooper, 2008, Insider trading before accounting scandals, Working
paper, University of Alabama.
Apfel, Robert C., John E. Parsons, G. William Schwert, and Geoffrey S. Stewart, 2001, Short sales,
damages, and class certifications in 10b-5 actions, Working paper, University of Rochester.
Asquith, Paul, and Lisa Meulbroek, 1996, An empirical investigation of short interest, Working
paper, Harvard University.
Asquith, Paul, Parag A. Pathak, and Jay R. Ritter, 2005, Short interest, institutional ownership,
and stock returns, Journal of Financial Economics 78, 243–276.
Bijwaard, Govert E., 2008, Instrumental variable estimation for duration data, Working paper,
Tinbergen Institute.
Cao, Bing, Dan S. Dhaliwal, and Adam C. Kolasinski, 2006, Bears and numbers: Investigating how
short sellers exploit and affect earnings-based pricing anomalies, Working paper, University
of Washington.
Christophe, Stephen E., Michael G. Ferri, and James J. Angel, 2004, Short-selling prior to earnings
announcements, Journal of Finance 59, 1845–1876.
Christophe, Stephen E., Michael G. Ferri, and Jim Hsieh, 2009, Informed trading before analyst
downgrades: Evidence from short sellers, Journal of Financial Economics 95, 85–106.
Cox, David Roxbee, 1972, Regression models and life tables, Journal of the Royal Statistical Society
Series B 34, 187–220.
Cox, James D., Randall S. Thomas, and Dana Kiku, 2003, SEC enforcement heuristics: An empirical
inquiry, Duke Law Journal 53, 737–779.
Daske, Holger, Scott A. Richardson, and Irem Tuna, 2005, Do short sale transactions precede bad
news events? Working paper, University of Pennsylvania.
D’Avolio, Gene M., 2002, The market for borrowing stock, Journal of Financial Economics 66,
271–306.
Dechow, Patricia M., Weili Ge, Chad Larson, and Richard Sloan, 2010, Predicting material accounting misstatements, Contemporary Accounting Research (forthcoming).
Dechow, Patricia M., Amy P. Hutton, Lisa Meulbroek, and Richard G. Sloan, 2001, Short-sellers,
fundamental analysis and stock returns, Journal of Financial Economics 61, 77–106.
Dechow, Patricia M., Richard G. Sloan, and Amy P. Sweeney, 1996, Causes and consequences
of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC,
Contemporary Accounting Research 13, 1–36.
Desai, Hemang, Srinivasan Krishnamurthy, and Kumar Venkataraman, 2006, Do short sellers
target firms with poor earnings quality? Evidence from earnings restatements, Review of
Accounting Studies 11, 71–90.
Desai, Hemang, K. Ramesh, S. Ramu Thiagarajan, and Bala V. Balachandran, 2002, An investigation of the information role of short interest in the NASDAQ market, Journal of Finance 57,
2263–2287.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
Short Sellers and Financial Misconduct
19:56
1913
Diether, Karl B., Kuan-Hui Lee, and Ingrid M. Werner, 2009, Short sale strategies and return
predictability, Review of Financial Studies 22, 575–607.
Diether, Karl B., Christopher J. Malloy, and Anna Scherbina, 2002, Differences of opinion and the
cross section of stock returns, The Journal of Finance 57, 2113–2141.
Duarte, Jefferson, Xiaoxia Lou, and Ronnie Sadka, 2006, Can liquidity events explain the lowshort-interest puzzle? Implications from the options market, Working paper, University of
Washington.
Dyck, Alexander, Adair Morse, and Luigi Zingales, 2010, Who blows the whistle on corporate
fraud? Journal of Finance (forthcoming).
Efendi, Jap, Michael Kinney, and Edward Swanson, 2006, Can short sellers anticipate accounting
restatements? Working paper, Texas A&M University.
Efendi, Jap, Anup Srivastava, and Edward. P. Swanson, 2007, Why do corporate managers misstate
financial statements? The role of option compensation and other factors, Journal of Financial
Economics 85, 667–708.
Einhorn, David, 2008, Fooling Some of the People All of the Time (John Wiley & Sons, Inc., Hoboken,
NJ).
Figlewski, Stephen, 1981, The informational effects of restrictions on short sales: Some empirical
evidence, Journal of Financial and Quantitative Analysis 16, 463–476.
Files, Rebecca, Edward P. Swanson, and Senyo Y. Tse, 2009, Stealth disclosure of accounting
restatements, Accounting Review (forthcoming).
Healy, Paul, 1985, The effects of bonus schemes on accounting decisions, Journal of Accounting
and Economics 7, 85–107.
Hennes, Karen M., Andrew J. Leone, and Brian P. Miller, 2008, The importance of distinguishing
errors from irregularities in restatement research: The case of restatements and CEO/CFO
turnover, The Accounting Review 83, 1487–1519.
Henry, Tyler R., and Jennifer L. Koski, 2010, Short selling around seasoned equity offerings,
Review of Financial Studies (forthcoming).
Jones, Charles M., and Owen A. Lamont, 2002, Short sale constraints and stock returns, Journal
of Financial Economics 66, 207–239.
Karpoff, Jonathan M., D. Scott Lee, and Gerald S. Martin, 2008a, The consequences to managers
for financial misrepresentation, Journal of Financial Economics 88, 193–215.
Karpoff, Jonathan M., D. Scott Lee, and Gerald S. Martin, 2008b, The cost to firms of cooking the
books, Journal of Financial and Quantitative Analysis 43, 581–612.
Karpoff, Jonathan M., D. Scott Lee, and Gerald S. Martin, 2009, The legal penalties for financial
misrepresentation: Preliminary evidence, Working paper. University of Washington and Texas
A&M University.
Lamont, Owen, 2004, Go down fighting: Short sellers vs. firms, NBER Working paper W10659.
Leinweber, David J., and Ananth N. Madhavan, 2001, Three hundred years of stock market
manipulations, The Journal of Investing 10, 7–16.
Liu, Ming, Tongshu Ma, and Yan Zhang, 2008, Are short sellers informed? New evidence from
short sales on financial firms during the recent subprime mortgage crisis, Working paper,
Binghamton University – SUNY.
Newey, Whitney K., and Kenneth D. West, 1987, A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55, 703–708.
Richardson, Scott A., 2003, Earnings quality and short sellers, Accounting Horizons (Supplement),
49–61.
Richardson, Scott A., Richard Sloan, Mark Soliman, and Irem Tuna, 2005, Accrual reliability,
earnings persistence and stock prices, Journal of Accounting and Economics 39, 437–485.
Wilchins, Dan, 2008, MBIA to urge curtailing short sellers, Reuters News Service, February 13,
2008.
Zarroli, Jim, 2008, SEC eyes stock manipulation in brokers’ collapse, NPR Business News, July
16, 2008.
jofi˙1597
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
jofi2009v2.cls (1994/07/13 v1.2u Standard LaTeX document class)
June 25, 2010
19:56
Queries
Q1 Author: Reference Cox (2008) has not been included in the reference list.
Please supply full publication details.
Q2 Author: Reference Diether et al. (2009) has not been included in the
reference list. Please supply full publication details.
1914